dcdbas driver question

Srinivas_G_Gowda at Dell.com Srinivas_G_Gowda at Dell.com
Tue Jan 19 03:26:35 CST 2016


Hi,
There seems to be no specific requirement on CPU# from which SMI needs to be invoked.
I am yet to get acknowledgment from few more folks.

Hope we don't break some unwritten Linux rule/best practice on _NOT_ sticking with CPU0 for issuing SMIs !!!!


ill probably do some testing with Juergen`s patch. Appreciate if anyone is able to test Juergen patch and share the results http://marc.info/?l=linux-kernel&m=145268587823715&w=2
Please note, this test is applicable only for Dell Systems and you would also need libsmbios to execute SMI based calls. I can help you out with the test procedure, please feel free to reach out to me.

@Juergen - Can you please share your test results as well.
b/w please keep libsmbios-devel at lists.us.dell.com CCd for any conversations based on dcdbas.


Thanks,
G


On Thursday 14 January 2016 03:04 PM, Srinivas_G_Gowda at Dell.com wrote:
> 
> Thanks Pali. I missed noticing that comment in the code. I cant find the CPU# restriction in any of the documents.
> Not sure why the condition was put there.... may be legacy!!! Not sure :(
> Let me run it through the BIOS folks and reconfirm.
> 
> 
> Thanks,
> G
> 
> 
> On Wednesday 13 January 2016 05:21 PM, Pali Rohár wrote:
>> On Wednesday 13 January 2016 17:03:05 Srinivas_G_Gowda at Dell.com wrote:
>>>>>> In the SUSE SLES Xen kernel we are carrying a patch for the dcdbas
>>>>>> driver to ensure a SMI is triggered on physical cpu 0 only. This
>>>>>> requires to use some Xen interfaces in the kernel which are not
>>>>>> part of the stable Xen API (this is the reason they have never been
>>>>>> upstreamed).
>>>>>>
>>>>>> We are now moving to a pvops kernel requiring to either port those
>>>>>> patches to pvops (and possibly upstream them) or to drop them in
>>>>>> case they are no longer needed.
>>>>>>
>>>>>> Are you able to tell me which Dell machines require a SMI to be
>>>>>> triggered on cpu 0? As my current Dell laptop is running with the
>>>>>> pvops kernel on top of Xen just fine, I guess not all Dell machines
>>>>>> have this requirement.
>>>>>>
>>>>>
>>>>> I believe there is no such restrictions. You should be able to get a valid response immaterial of which CPU issues the SMI.
>>>>
>>>> So dcdbas_smi_request() should be changed to not bind the running
>>>> thread to cpu 0? In case you confirm, I can set up a patch doing
>>>> this.
>>>>
>>> Yes, you don't need to have an affinity for CPU0.
>>
>> Really? Code in commit 90563ec4129f14d19f018240d1d3ff5c0e5e6392
>> of linux git tree explicitly says that "SMI requires CPU 0"...
>>
> _______________________________________________
> libsmbios-devel mailing list
> libsmbios-devel at lists.us.dell.com
> https://lists.us.dell.com/mailman/listinfo/libsmbios-devel
> 


More information about the libsmbios-devel mailing list